[Python-Dev] Behavior change in subprocess.py

[Kevin Jacobs <jacobs@bioinformed.com>]

During my testing of Python 2.5b2, I've found something that may be worthy
of discussion.  I suspect that recent GC and finalization changes have
altered the behavior of the Popen object in subprocess.py.  I am now getting
many many many finalization warnings in my code like:

Exception exceptions.AttributeError: "'NoneType' object has no attribute
'append'" in <bound method Popen.__del__ of <subprocess.Popen object at
0x2aaaab910950>> ignored

Is this a bug or a feature?  Personally, I'd like to see these messages
silenced, since it is being generated during interpreter shutdown.  The
following patch does the trick for me:

--- /usr/local/lib/python2.5/subprocess.py      2006-07-11 14:11:
59.000000000 -0400
+++ subprocess.py       2006-07-12 10:17:09.000000000 -0400
@@ -613,7 +613,7 @@
             return
         # In case the child hasn't been waited on, check if it's done.
         self.poll(_deadstate=sys.maxint)
-        if self.returncode is None:
+        if self.returncode is None and _active is not None:
             # Child is still running, keep us alive until we can wait on
it.
             _active.append(self)


Note that popen.py does something similar, though I am not convinced that
the test is right or if it is doing something more subtle:

    def __del__(self):
        # In case the child hasn't been waited on, check if it's done.
        self.poll(_deadstate=sys.maxint)
        if self.sts < 0:
            if _active:
                # Child is still running, keep us alive until we can wait on
it.
                _active.append(self)

Regards,
-Kevin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.python.org/pipermail/python-dev/attachments/20060712/6b697648/attachment.htm